Evolutionary Computation for Sparse Multi-Objective Optimization: A Survey
Shuai Shao, Ye Tian, Yajie Zhang, Shangshang Yang, Panpan Zhang, Cheng He, Xingyi Zhang, Yaochu Jin
Abstract
In various scientific and engineering domains, optimization problems often feature multiple objectives and sparse optimal solutions, which are commonly known as sparse multi-objective optimization problems (SMOPs). Since many SMOPs are pursued based on large datasets, they involve a large number of decision variables, leading to a huge search space that is challenging to find sparse Pareto optimal solutions. To address this issue, a number of multi-objective evolutionary algorithms (MOEAs) have been developed in recent years to identify non-zero variables through novel search strategies. However, there is currently limited literature that systematically reviews the related studies. In this article, a comprehensive survey is presented for sparse multi-objective optimization, which starts with a definition of SMOPs, followed by a taxonomy of existing sparse MOEAs. Then, the sparse MOEAs are reviewed in detail, followed by an introduction of benchmark and real-world applications that are used for performance assessment in sparse optimization. Finally, the survey is finished by summarizing the research status of sparse multi-objective optimization and outlining some promising research directions.